About
I am a Full Professor in the Department of Computer Systems and Computation at the Universitat Politècnica de València, and a member of the DMIP group at VRAIN, the Valencian Research Institute for Artificial Intelligence, where I have done research since 1999. I am also a member of ValgrAI, the Valencian Graduate School and Research Network of Artificial Intelligence. Since 2009 I have been Assistant Dean for University–Enterprise Relations at the School of Computer Science (ETSINF). I have also directed the Chair in Artificial Intelligence Applied to Public Administration (Cátedra de Inteligencia Artificial aplicada a la Administración Pública) since 2021, and the university extension diploma in Artificial Intelligence of the Samsung Innovation Campus since 2022.
My research began with how to evaluate classifiers properly: ROC analysis, AUC, calibration, cost curves and threshold choice. It has since grown into a broader question: how to evaluate and predict the behaviour of AI systems, from classifiers to large language models and autonomous agents. I also apply machine learning to real-world problems, including urban mobility, air quality, environmental monitoring and education, and I maintain several open, continuously updated datasets.
I received an MSc in Computer Science (1998) and a PhD (2003) from the Universitat Politècnica de València. My thesis was on applying declarative languages to machine learning.
News
- Talk “Los riesgos reales de la IA: de los sesgos a los ataques” at the Ateneo Mercantil de València (14 October 2026).
- Presented “Size-constrained clustering via bat algorithm optimization” (with D. Vallejo-Huanga, A. Padilla and F. Martínez-Plumed) at Discovery Science 2026, Mainz, Germany (October 2026).
- Interviewed by Newtral: “¿Por qué las IA hacen más trampas? ‘Es el camino más directo’” (6 October 2026).
- Attended ECML PKDD 2026 in Naples, Italy (September 2026).
- Member of Spain's Open Government Forum (Foro de Gobierno Abierto) since September 2026.
- Organiser of IJCLR 2026, the International Joint Conference on Learning & Reasoning.
- Research stay at the Leverhulme Centre for the Future of Intelligence (CFI), University of Cambridge, June–July 2026.
- Talk on the Salvem les Fotos UPV project at the University of Cambridge (June 2026).
- Talk “La IA generativa y los modelos del lenguaje: GPT, Gemini, Llama” at the Museu de les Ciències, Ciutat de les Arts i les Ciències, València (3 March 2026) · video.
- Talk on the Salvem les Fotos UPV project at CEU University (March 2026).
- Member of Salvem les Fotos UPV / Recuperar las memorias since November 2024, the UPV project to recover family photographs damaged by the 2024 València floods. Our AI work for the project appeared at AAAI 2026: “AI for memory preservation: Automated restoration of photographs damaged by floods”.
- “When redundancy matters: Machine teaching of representations” is published in Machine Learning (2026).
- “Predictable artificial intelligence” is published in Artificial Intelligence (2026).
Research interests
ML evaluation
ROC analysis, AUC, performance metrics, calibration, cost curves and threshold choice.
Predictable & reliable AI
Measuring and anticipating the capabilities and failures of LLMs and AI agents.
Quantification
Estimating class prevalence under dataset shift for classification and regression.
Explainability & machine teaching
Interpretable models, XAI for time series, teaching size and inductive programming.
Data science processes
CRISP-DM and data science trajectories, automated data wrangling, size-constrained clustering.
Applied AI
Urban mobility and pollution, microfibre detection in wastewater, photo restoration, learning analytics.
Selected publications
- Larger and more instructable language models become less reliable
Nature, 2024 - Predictable artificial intelligence
Artificial Intelligence, 2026 - Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models
Transactions on Machine Learning Research, 2023 - CRISP-DM twenty years later: From data mining processes to data science trajectories
IEEE Transactions on Knowledge and Data Engineering, 2019 - Can language models automate data wrangling?
Machine Learning, 2022 - A unified view of performance metrics: translating threshold choice into expected classification loss
Journal of Machine Learning Research, 2012 - An experimental comparison of performance measures for classification
Pattern Recognition Letters, 2009 - Learning decision trees using the area under the ROC curve
International Conference on Machine Learning (ICML), 2002
Software & open data
- València urban-mobility archive Continuously collected open data from València: Valenbisi bike-sharing, traffic intensity, loop detectors and parking occupancy (2015–today).
- JCDecaux GBFS archive Daily archive of bike-share station status for all 19 cities operated by JCDecaux/Cyclocity.
- Agent-trace dataset Agent-execution traces (Inspect AI) of LLM agents across coding and tool-use benchmarks.
- ADeLe annotation distillation Multi-model demand-level annotation and a distilled 7B annotator for AI evaluation.
- BSPA & EMQSPA Competitive quantification at low computational cost.
- Microfibre segmentation quality Label-free estimation of instance-segmentation quality for microfibre detection.
- Cost curves R tools for cost curves, ROC analysis and threshold-choice methods.
Earlier software
- FLIP A functional logic inductive programming system.
- SMILES A Stunning Multi-purpose Integrated LEarning System.
- SYNTH & CLOP A transformation system and an automatic optimiser for functional logic programs.
- Brier curves, rate-driven curves & Reg2Class R scripts for threshold choice in cost space (2011–2012) and for turning regression into classification (2014).
- Weka page My Weka resources.
I co-authored Introducción a la minería de datos, one of the first books on data mining in Spanish.
PhD students
Current
Completed theses
- Daniel Silva-Palacios (2021) Clasificación jerárquica multiclase
- Lidia Contreras-Ochando (2020) Towards data wrangling automation through dynamically-selected background knowledge
- Fernando Martínez-Plumed (2016) Incremental and developmental perspectives for general-purpose learning systems
- Antonio Bella (2012) Model integration in data mining: From local to global decisions
Industry & public-sector collaborations
Service & teaching
Editorial boards
Machine Learning · Engineering Applications of Artificial Intelligence · Revista Iberoamericana de Inteligencia Artificial
Programme committees
NeurIPS, ICLR, ICML, IJCAI, AAAI, KDD, ECML-PKDD, WI, MOD, CBMS, AI, IDEAS, ICMLA, IAT and others.
Teaching
My current subjects and office hours are listed on my UPV staff page.
Contact
Office D235 · Department of Computer Systems and Computation (DSIC)
Universitat Politècnica de València
Camí de Vera s/n, 46022 València, Spain
✉ cferri@dsic.upv.es ·
☎ +34 963 877 000 (ext. 83505)
Also on ResearchGate · Scopus · ResearcherID